Since the Health Information Technology for Economic and Clinical Health (HITECH) Act was enacted, provider notes have lengthened with an increase in note redundancy [1]. Across the same period, US notes have been reported as 4 times longer than those in other countries [2]. Some claim that poor note quality is one of the unintended consequences of nationwide electronic health record (EHR) adoption [3-6]. The introduction of electronic workflows has allowed the flow of raw uncontextualized clinical data into provider notes [5,7-9]. Over time, and with the change in the culture of clinical practice, many electronic notes have become full of noise, and clinically significant signals are now  becoming harder to find [1,3,5,9,10]. There is an urgent need to develop and implement innovative approaches in electronic clinical documentation that improve note quality and reduce unnecessary bloating [5,8,9].

Significant unintended consequences to care associated with the adoption of EHR are rooted in unplanned changes to the sequence of complex clinical events. EHR interventions can have unintended changes to workflow, which is defined as the sequence of activities necessary to complete a task, and patient flow, which is described as the ordered movement of patients in a health care setting. Examples of disordered workflow after EHR adoption include changes to sequence of complex physician ordering processes after the implementation of computerized physician order entry [11] and changes to temporal trends in physician task workflow as well as increased time spent in clinical review and documentation [12]. Some examples of changes in patient flow included a lack of understanding of the sequence of patient event after transition from paper to electronic flowsheets [13] and disordered patient flow with the requirement that staff need physician orders prior to initiating “routine” clinical testing [14].

Sequence also plays a role in clinical thinking and communication processes. SBAR (situation, background, assessment, and recommendation) [15], IPASS (illness severity, patient summary, action list, situation awareness, and contingency planning and synthesis by receiver) [16], and SOAP (subjective, objective, assessment, and plan) [17] are frameworks, or mental models, that advance ordered standards for thought and communication. SBAR and IPASS are used at the time of clinical handoff, and SOAP is a construct for clinical notes. These standards promote a sequential structure where adherence to sequence is a key element to improve patient safety and impact other clinical outcomes [18-20]. In this context, the Data, Information Knowledge, Wisdom (DIKW) framework is a valuable tool that has been used to study clinical communication [21-24]. The framework is composed of 4 levels, each building on the prior. Data represent the initial stage, where raw facts such as patient vital signs and laboratory results are collected. Information is the next stage, where data such as vitals trend are organized and processed. Knowledge is the third stage, where information such as diagnosis or assessment is interpreted and applied. Wisdom is the final stage, where knowledge such as considering a patient’s cultural and personal values when making a treatment plan is applied in a practical and integrated manner. By using the DIKW framework, researchers can analyze the flow of communication and identify any gaps or inefficiencies in the sequential progression from data to wisdom [25-30].

Considering the importance of sequence and flow to clinical thinking, clinical communication, and EHR workflows in general, it is not surprising that electronic note-writing introduced features that change the flow of information and can impact the output of the note composition. The original SOAP note structure, created by Larry Weed, was intended as a data transformation engine that takes raw clinical data and transforms it into information in the “Subjective” and “Objective” sections [31]. For instance, the patient may report a headache and difficulty sleeping (subjective), and the clinician may find that the patient’s vital signs are within normal limits but with tenderness in the temples upon physical examination (Objective). This information is then converted into knowledge in the “Assessment” section, such as the conclusion that the patient is likely experiencing a headache due to stress and lack of sleep. Finally, the knowledge is transformed into wisdom in the “Plan” section, with a recommendation for over-the-counter pain relief medication and advice for the patient to practice stress management techniques and establish a consistent sleep schedule. In this way, the note writer sequentially [32,33] ascends the DIKW pyramid [21], while the SOAP note is composed (Figure 1). Modern electronic note-writing workflows have disrupted this time-tested process for data transformation by allowing data and information to flow directly from the EHR into predefined note sections before the clinician thinks about the patient’s condition. Untimely information flow into note templates can thus be viewed as a root cause of poor note quality and a major contributor to the note bloat epidemic.

The objective of this study is to describe the development and impact of a novel set of templates designed to change the flow of information into provider notes (Figure 2).

We describe the development process, including a description of institutional governance, guiding principles that informed template design, template creation, usability testing, and implementation.

Usability testing is essential to the EHR implementation life cycle [34-38], and testing events can help promote provider adoption [39,40]. Combining “near-live” clinical simulation with “think-aloud” protocols has been shown to help assess user preference and impact on workflow [41]. “Think-aloud” protocols require users to verbalize their thought process while interacting with a new clinical tool [41-44]. “Near-live” testing allows for a more fluid environment to identify further real-life barriers [34,41]. “Near-live” clinical simulation with “think-aloud” protocols can be deployed together with quantitative usability assessment like the System Usability Scale (SUS) [45-47] as part of a pragmatic usability test strategy [34] that prioritizes speed and cost-effectiveness. The SUS has been used extensively to assess the usability of the EHR [42,48-50], including use to assess electronic note template usability [51]. The 10 statements on the SUS use a 5-point Likert scale to measure the strength of agreement or disagreement with each statement. The total composite score is a number from 0 to 100, with high SUS scores indicating greater usability and satisfaction.

This is a multihospital nonrandomized prospective improvement study conducted on the inpatient general internal medicine (GIM) service across 3 hospital campuses at the New York University Langone Health System (NYULH). NYULH is a large academic health care system in New York, consisting of over 5000 health care providers across multiple hospitals and >500 ambulatory locations. NYULH’s implementation of a single EHR (Epic Systems Corporation) instance and seamless integration of ancillary systems help to facilitate a single clinical standard throughout the enterprise. To support measurement and continuous improvement for this project, EHR data were queried from the Epic System’s Clarity database using SQL Developer (Oracle Corporation) and exported for analysis of note length and template adoption. The note templates were implemented in the November 2020 version of Epic. Prior to our quality improvement intervention, the institution’s instance of Epic does not allow “copy forward” in the inpatient setting. NYULH did not have any standardized note templates prior to our intervention.

We followed the NYU Grossman School of Medicine institutional review board (IRB) protocol and completed an IRB checklist for activities that may be classified as quality improvement. This work, including our usability study, met the IRB criteria for quality improvement and therefore did not require IRB review or informed consent.

Our documentation improvement efforts began in September 2020, when a group of physician leaders from NYULH representing each hospital campus began to meet biweekly as part of the newly formed Documentation Standards Committee. The committee’s stated goal was to develop and implement standardized documentation and accountability processes for provider notes to improve quality and readability, while eliminating errors, and capturing the complexity of care provided. Figure 3 represents the timeline for the improvement initiative and committee structure, respectively. Figure 4 is a list of key committee stakeholders.

In the first 6 months, during the committee meetings, we conceptualized note bloat. As a result of these committee meetings, we introduced the bundled intervention that included (1) conceptualization of the note bloat problem as a dysfunction in information flow (Figures 1 and 2); (2) a set of guiding principles for organizational documentation improvement (Figure 5); (3) a documentation improvement curriculum for inpatient medicine providers; and (4) the design and build of novel new electronic templates that sought to reduce the flow of extraneous information into provider notes by providing link outs to best practice data visualizations.

Figure 6 shows the progress note template developed for the inpatient GIM service. The first implementation of new standard templates was for GIM and included: progress notes, consults, and history and physical (H&P) notes. Within these new templates, very little clinical data flow directly into the note. Instead, each template section has hyperlinks that link out to best visualizations that appear next to the template to support clinical data review. These hyperlinks provide direct access to data-specific sections within the patients’ medical record including medical and surgical history; for example, the data are visualized in a side panel next to the open note. This allows for clinical data review and simultaneous entry of relevant details into the clinical note. Providers are prompted within the template to document a synthesis of the information or knowledge relevant to the note section. Each note section contains a unique combination of hyperlinks with tailored data visualizations to support appropriate clinical thinking in that part of the note. All the hyperlinks disappear at the time of note publication.

The templates generally follow the SOAP note structure, although additional sections were added for clarity or to adapt to the modern practice of medicine. The template sections are titled as follows: Subjective & Notable Events, Physical Exam, Laboratory Test & Imaging Review, Assessment & Plan, and Discharge Milestones & Contingency Planning. This last section continues the sequential progression of the SOAP note structure by moving onto future and contingency planning, including how the patient's care is affected by health system considerations about the advancement of care and hospital discharge.

An additional feature of these novel templates is the presence of patient-specific CDS presenting as disappearing tips that appear in the body of the note. Figure 7 demonstrates a CDS nudge directed at providers to consider the plan for anticoagulation in patients who are on full-dose anticoagulation up to 48 hours prior to surgery. Figure 8 is an example of another CDS nudge to providers to consider the plan for antibiotics, upon discharge, for patients who are currently on intravenous antibiotics. The usability of CDS nudges within note sections was evaluated as part of formal usability testing as described below.

Prior to go-live, pragmatic usability testing was performed with the new templates. Four board-certified physician informaticists served as facilitators for 45-minute one-on-one testing sessions with frontline providers. A purposive sample of inpatient internal medicine providers was selected: 11 from NYU Langone—Long Island Hospital, 3 from NYU Langone—Brooklyn Hospital, and 8 from the Tisch and Kimmel Hospitals on the Manhattan campus. Half of the providers were attending physicians and half were residents or physician assistants. All the providers routinely write notes on hospitalized general medicine service patients.

The templates and hyperlinks to data visualizations were built within the sandbox testing environment of the EHR to function the same as when live in the production system. The testing protocol combined “near-live” clinical simulation of the note-writing processes with “think-aloud” protocol and task-based testing. In this pragmatic usability test, physician informaticists collected data including observational notes, notes on debriefing interviews, and real-time analysis of user-screen interactions. At the end of each testing session, participants completed the 10-question SUS. Summative analysis included a structured summary of themes from physician informaticists debriefing and session notes, average SUS scores, and response distribution by question for the 22 session participants. Themes and SUS scores were presented to the governance committee, and improvements were made to the template prior to go-live (Figure 9).

Two weeks prior to go-live, all 350 internal medicine providers in the institution were assigned an electronic learning module with an introductory message from the committee chair, elucidation of guiding principles for documentation improvement, and demonstration of the new template in the electronic record. In-person roadshows were conducted in more than 15 academic conferences, quality forums, department, and division meetings across the 4 hospital campuses. The template went live on March 8, 2021. Post live, physician informaticists rounded on all 3 hospital campuses for the first week to support adoption and answer questions about the new approach to electronic documentation.

Template utilization and note length in characters were the primary outcomes measured. We used the month of November prior to provider outreach as the baseline preintervention data and the 3 months immediately following implementation in March as the postintervention data.